CN102939628B - Apparatus and method for processing an input audio signal using cascaded filterbanks - Google Patents

Abstract

An apparatus for processing an input audio signal (2300) relies on a cascade of filterbanks, the cascade comprising a synthesis filterbank (2304) for synthesizing an audio intermediate signal (2306) from the input audio signal (2300), the input audio signal being represented by a plurality of first subband signals (2303) generated by an analysis filterbank (2302), wherein a number of filterbank channels of the synthesis filterbank (2304) is smaller than a number of channels of the analysis filterbank (2302). The apparatus furthermore comprises a further analysis filterbank (2307) for generating a plurality of second subband signals (2308) from the audio intermediate signal (2306), wherein the further analysis filterbank has a number of channels being different from the number of channels of the synthesis filterbank (2304), so that a sampling rate of a subband signal of the plurality of second subband signals (2308) is different from a sampling rate of a first subband signal of the plurality of first subband signals (2303).

Description

In order to use the device and method of cascading filter group process input audio signal

Technical field

The present invention relates to audio source coding systems, this system utilizes one for the harmonic wave transposition method of high-frequency reconstruction (HFR), and relate to digital effect processor, such as so-called driver, wherein the generation of harmonic distortion adds the brightness of treated signal, and relate to time explanation device, wherein the duration of signal is extended and maintains original spectral content simultaneously.

Background technology

In PCT WO 98/57436, the concept of transposition is established a kind of method producing a high frequency band as a low-frequency band from a sound signal again.By a large amount of saving using this concept can obtain bit rate in audio coding.In an audio coding system based on HFR, by core wave coder process one low-bandwidth signal, and the extremely low bit rate of utilization to the target spectrum shape describing decoder-side carries out transposition and adds side information to regenerate upper frequency.For low bit rate, when the narrow bandwidth of core encoder signal, the high frequency band that regeneration one has the joyful characteristic of perception becomes more and more important.In PCT WO 98/57436, the harmonic wave transposition of definition performs very well to complicated music material in the situation with low crossover frequency.The principle of harmonic wave transposition is the sinusoidal curve that a sinusoidal curve with frequencies omega is mapped to that has frequency T ω, and wherein T > 1 is the integer on definition transposition rank.In contrast to this, the HFR method based on single-sideband modulation (SSB) being has frequencies omega sinusoidal curve by one is mapped to the sinusoidal curve that has the frequency of frequencies omega+△ ω, and wherein △ ω is a fixing frequency displacement.Suppose that a core signal has low bandwidth, SSB transposition may cause the inconsonant ring artifact of generation one.

In order to reach audio quality as well as possible, up-to-date high-quality harmonic wave HFR method uses complicated modulated filter bank, such as short time Fourier transform (STFT), reaches the audio quality of expectation with high frequency resolution and a height over-sampling.Need fine-resolution to avoid because not wanting intermodulation distortion caused by Nonlinear Processing sinusoidal curve summation.When enough high frequency resolution, namely narrow sub-band, high-quality method object is to make to have a sinusoidal curve maximal point in each sub-band.Need temporal height over-sampling to avoid the distortion of aliasing type, and need the to a certain degree over-sampling in frequency to avoid the Pre echoes of momentary signal.Obvious deficiency is that the complexity calculated can uprise.

Harmonic wave transposition based on sub-band block is another HFR method for suppressing intermodulation products, and in the case, employing one has the bank of filters compared with coarse frequency resolution and a low degree over-sampling, a such as hyperchannel QMF group.In the method, by a common phase modifier process, the superposition of several adjustment sample forms an output subband samples to the time block of a multiple subband samples.This has the net effect suppressing intermodulation products, otherwise this intermodulation products will there will be when inputting sub-band signal and being made up of several sinusoidal curve.More much lower than high-quality transposition device and almost identical quality is obtained to many signals on computation complexity based on the transposition of the sub-band process based on block.But complexity is still far above the common HFR method based on SSB, this is because need multiple analysis filterbank in a typical HFR application, the signal of the different transposition rank T of each bank of filters process, to synthesize the bandwidth of needs.In addition, a common mode is the analysis filterbank making the sampling rate adaptation one of input signal have fixed size, although the signal on the different transposition rank of bank of filters process.Meanwhile, also belong to and commonly use bandpass filter to obtain via the process of different transposition rank, the output signal with non-overlapped power spectrum density to input signal.

The storage of sound signal or transmission are subject to strict bit rate restriction often.In the past, when only having a low-down bit rate to utilize, scrambler is forced to the audio bandwidth significantly reducing transmission.Modern audio codecs is now by utilized bandwidth expansion (BWE) method [1 ?12] encoded bandwidth signal.These algorithms rely on radio-frequency component (HF) a Parametric Representation, this radio-frequency component be by transposition to (" repairings ") in HF spectral regions and apply a driving parameter aftertreatment generate from the low frequency part (LF) of decoded signal.LF part is by any audio frequency or speech coder coding.For example, the bandwidth expanding method described in [1 ?4] relies on single-sideband modulation (SSB), usually also referred to as " copying " method, to produce multiple HF patch.

Recently, a kind of new algorithm of the phase vocoder group [15 ?17] for generation of different patch that uses is suggested [13] (see Figure 20).This method has been developed for avoiding the sense of hearing coarse, and the sense of hearing is coarse to be observed usually on the signal through SSB bandwidth expansion.But, because BWE algorithm performs at the decoder-side of a codec chain, so the complexity calculated is serious problems.State-of-the-art method, especially based on the HBE of phase vocoder, with compared with SSB method, is for obtaining under cost with a computation complexity greatly increased.

As outlined above, existing bandwidth extension schemes only applies a method for repairing and mending to a given signal block, and it is the repairing [1 ?4] based on SSB or the repairing [15 ?17] based on HBE vocoder.In addition, modern audio codec [19 ?20] provides and switches the possibility of method for repairing and mending based on time block overall situation between selectable mending option.

SSB copies and forms patch unnecessary roughness is introduced sound signal, but calculates simple and remain the temporal envelope of transient state.In addition, computation complexity increases significantly and exceedes the very simple SSB clone method of calculating.

Summary of the invention

When complexity of touching upon reduces, sampling rate has special importance.This is because a high sampling rate means high complexity, and a low sampling rate means to have low complex degree usually due to required operation decreased number.But, on the other hand, bandwidth expansion application situation especially true and core encoder is outputed signal sampling rate by typically the low sampling rate to making a full bandwidth signal is too low.State by different way, when the sampling rate of decoder output signal be such as 2 or 2.5 be multiplied by the maximum frequency of core encoder output signal time, then a such as factor be 2 bandwidth expansion mean that needs one up-sampling operates, make the sampling rate of bandwidth expansion sampled signal high to making sampling that the radio-frequency component produced in addition " can be contained ".

In addition, the bank of filters of such as analysis filterbank and synthesis filter banks, is responsible for quite a large amount of process operations.Therefore, the size of bank of filters, namely bank of filters be one 32 path filter groups, one 64 path filter groups or the bank of filters of even higher number will affect the complexity of audio processing algorithms significantly.Usually can say, the filter bank channel of a high number needs more process operation, and thus less than number filter bank channel complexity is high.In view of this, bandwidth expansion application and in different sampling rates be also a key other audio frequency process application (such as in the application of similar vocoder or other audio frequency effect any application) in, there is specific a depending on each other for existence property between complexity and sampling rate or audio bandwidth, mean when selecting instrument or the algorithm of mistake for specific operation, the operation of up-sampling or sub-band filtering significantly increases complexity before not affecting audio quality especially on positive sense.

An object of the present invention is to provide the audio frequency process concept of an improvement, and this audio frequency process concept allows low complex degree process on the one hand, obtains a good audio quality on the other hand.

This object is realized by the apparatus and method according to following record.

The device of a kind of process one input audio signal is provided, comprise: a synthesis filter banks, in order to synthesize an audio frequency M signal from described input audio signal, multiple first sub-band signals that described input audio signal is produced by an analysis filterbank represent, the filter bank channel number of wherein said synthesis filter banks is less than the number of active lanes of described analysis filterbank; And another analysis filterbank, in order to produce multiple second sub-band signal from described audio frequency M signal, the number of active lanes that another analysis filterbank wherein said has is different from the number of active lanes of described synthesis filter banks, makes the sampling rate of the sub-band signal in described multiple second sub-band signal different from the sampling rate of one first sub-band signal in described multiple first sub-band signal.

The device of a kind of process one input audio signal is also provided, comprise: an analysis filterbank, there is multiple analysis filterbank passage, wherein, described analysis filterbank is configured to carry out filtering to described input audio signal, to obtain multiple first sub-band signal; And a synthesis filter banks, an audio frequency M signal is combined into for what use the first sub-band signal, wherein, the described group of sub-band signal comprising a number and be less than the filter bank channel number of described analysis filterbank, wherein said middle sound signal is that the sub sampling of a portions of bandwidth of described input audio signal represents.

The method of a kind of process one input audio signal is provided, comprise: use the synthesis filter banks being used for synthesizing from described input audio signal an audio frequency M signal to carry out synthetic filtering to the described input audio signal that multiple first sub-band signals produced by an analysis filterbank represent, wherein, the number of active lanes of the bank of filters of described synthesis filter banks is less than the number of active lanes of described analysis filterbank; And use another analysis filterbank being used for producing from described audio frequency M signal multiple second sub-band signal to carry out analysis filtered, the number of active lanes that another analysis filterbank wherein said has is different from the number of active lanes of described synthesis filter banks, makes the sampling rate of a sub-band signal of described multiple second sub-band signal be different from the sampling rate of one first sub-band signal of described multiple first sub-band signal.

A kind of method for the treatment of an input audio signal, comprise: use an analysis filterbank with multiple analysis filterbank passage to carry out analysis filtered, wherein said analysis filterbank is configured to input audio signal described in filtering to obtain multiple first sub-band signal; And use for utilizing the synthesis filter banks being combined into an audio frequency M signal of the first sub-band signal to carry out synthetic filtering, wherein, the number of the described group of sub-band signal comprised is less than the filter bank channel number of described analysis filterbank, wherein, described middle sound signal is that the sub sampling of a portions of bandwidth of described input audio signal represents.

The specific cascade layout of embodiments of the present invention dependency analysis and/or synthesis filter banks obtains low complex degree resampling not sacrificing under audio quality.In one embodiment, the device processing an input audio signal comprises a synthesis filter banks, in order to synthesize an audio frequency M signal from input audio signal, wherein this input audio signal is that wherein the filter bank channel number of this synthesis filter banks is less than the number of active lanes of this analysis filterbank to be represented by multiple first sub-band signals that produce of analysis filterbank before is placed in composite filter on process direction.M signal is further by another analysis filterbank process for producing multiple second sub-band signal from this audio frequency M signal, wherein the number of active lanes of this another analysis filterbank is different from the number of active lanes of this synthesis filter banks, makes the sampling rate of the sub-band signal in the plurality of sub-band signal be different from the sampling rate of the first sub-band signal in multiple first sub-band signals produced by this analysis filterbank.

One synthesis filter banks and one provides a sample rate conversion with the cascade of another analysis filterbank latter linked, and provides the modulation of portions of bandwidth to base band of the original audio input signal having inputted synthesis filter banks in addition.Present extraction is preferably expressed as from this time M signal of original input audio signal the threshold sampling signal being modulated to base band at present, this original input audio signal such as can be the output signal of a core decoder of a bandwidth extension schemes, and found this expression (namely, this resampling outputs signal) when being represented to obtain a sub-band by another analysis filterbank process, allow the low complex degree process of process operation further, this further process operation may or may can not occur, and this process operation that process operational example is relevant to bandwidth expansion in this way further, the merging of the high-frequency reconstruction process after such as nonlinearities frequency band operation and in the end synthesis filter banks sub-bands.

The different aspect of the device of audio signal, method or computer program under the application is provided in the background of bandwidth expansion and under other voice applications background had nothing to do with bandwidth expansion.Then the feature described and claimed various aspects can partly or entirely merge; but also can use independently of each other, this is because when various aspects are implemented in a computer system or microprocessor, provide the advantage about perceptual quality, computation complexity and processor/memory resource.

Embodiment provides a kind of method, and the method is in order to reduce one based on the computation complexity of the harmonic wave HFR method of sub-band block by carrying out efficient filtering and sample rate conversion to the input signal to HFR analysis filterbank AG.In addition, the bandpass filter being applied to input signal can be shown in one based on being useless in the transposition device of sub-band block.

Present embodiment by implementing the computation complexity promoting the harmonic wave transposition reduced based on sub-band block based on several rank of the transposition of sub-band efficiently in a single analysis and the right framework of synthesis filter banks.The perceptually shifting relation of quality and computation complexity, can the place of execution only suitable subset on the rank of transposition or all rank of transposition jointly in a bank of filters.In addition, in the transposition scheme of a combination only some transposition rank by directly calculate and remaining bandwidth be by available (that is, the transposition rank (such as second-order) previously calculated and/or core encoder bandwidth copy filling.In the case, each that can use the available source range for copying may combine and performs repairing.

In addition, embodiment provides a kind of method, and the method improves high-quality harmonic wave HFR method and the harmonic wave HFR method based on sub-band block by the spectral alignment of HFR instrument.Particularly, aim at realize performance enhancement by the adjust frequency spectral boundaries shown of the spectral boundaries that HFR produced signal and envelope.In addition, the spectral boundaries of limiter instrument is aligned to identity principle the spectral boundaries that HFR produces signal.

Further embodiment is configured to for improving instantaneous perceptual quality and such as reduce computation complexity by application one mending option simultaneously, and the application of this mending option is repaired by harmonic wave and copied mixing that repairing forms and repair.

In specific embodiment, each bank of filters of cascading filter group structure is quadrature mirror filter bank (QMF), all lowpass prototype filter or windows all depending on the set modulation of the modulating frequency of the centre frequency using definition filter bank channel.Preferably, all window functions or prototype filter are depending therefrom in the mode having wave filter in the bank of filters of different size (filter bank channel) also depending therefrom.Preferably, maximal filter group in bank of filters cascade construction comprises one first analysis filterbank, in one embodiment with latter linked bank of filters, another analysis filterbank and the last synthesis filter banks after a while in treatment state, and this synthesis filter banks comprises the window function or prototype filter response with given number window function or prototype filter coefficient.The bank of filters of reduced size is all the sub-sampled version of this window function, and the window function meaning other bank of filters is the sub-sampled version of " greatly " window function.For example, if a bank of filters has the half size of large bank of filters, then window function has the coefficient of half number, and the coefficient of the less bank of filters of size is obtained by sub sampling.In this case, sub sampling means such as each second filter coefficient and is regarded as having the less bank of filters of half size.But when having other relation between non-integral bank of filters size, then perform certain interpolation of window coefficient, make last, the window of less bank of filters is the sub-sampled version of the window of larger bank of filters again.

Embodiments of the present invention are particularly useful when process needs an only part for input audio signal further, and this situation especially occurs under the occasion of harmonic wave bandwidth expansion.In this occasion, the process operation of vocoder and so on is particularly preferred.

An advantage of embodiment is that embodiment provides by efficient time-domain and frequency-domain operation the complexity that a QMF transposition device is lower, and utilizes spectral alignment to be the audio quality providing improvement based on the harmonic spectrum tape copy of QMF and DFT.

Embodiment relates to a kind of audio source coding systems, this system uses and is such as used for high-frequency reconstruction (HFR) based on the harmonic wave transposition method of sub-band block, and relate to digital effect processor (such as so-called driver), wherein the generation of harmonic distortion adds the brightness of processing signals, and relate to time explanation device, wherein the duration of signal is extended and keeps original spectrum component simultaneously.Embodiment provides a kind of by carrying out to input signal the method that high-efficient filter involves sample rate conversion to reduce the computation complexity of the harmonic wave HFR method based on sub-band block before HFR filter bank analysis level.Further, embodiment shows the conventional band-pass filters being applied to input signal is useless in a HFR system based on sub-band block.Additionally, embodiment provides a kind of method, and the method improves high-quality harmonic wave HFR method and the harmonic wave HFR method based on sub-band block by the spectral alignment of HFR instrument.Particularly, embodiment teaches and how aims at by the spectral boundaries of signal that produced by HFR and the envelope spectral boundaries shown of adjusting frequency the enhancing realizing performance.Further, the spectral boundaries of limiter instrument is aligned to HFR with identical principle to produce the spectral boundaries of signal.

Embodiment

Following embodiment is only illustrative, and provides a low complex degree of QMF transposition device by efficient time domain and frequency-domain operations, and is provided the audio quality of improvement of the harmonic wave SBR based on QMF and DFT by spectral alignment.Will be understood that, amendment described herein and configuration variation and details are apparent for those skilled in the art.Therefore be only limited to claim scope and be not limited to by the description of embodiment herein with the specific detail proposed is described.

Figure 23 shows preferably the implementing of device for the treatment of an input audio signal, and wherein input audio signal can be the time domain input signal that a such as core audio demoder 2301 online 2300 exports.Input audio signal is input to one first analysis filterbank 2302, and it is such as an analysis filterbank with M passage.In detail, therefore analysis filterbank 2302 exports M sub-band signal 2303, and its sampling rate had is fS=fS/M.This means analysis filterbank is a threshold sampling analysis filterbank.This means for each block be made up of M input amendment on line 2300, and analysis filterbank 2302 provides each sub-band passage for single sample.Preferably, analysis filterbank 2302 is bank of filters of a multiple modulation, means each subband samples and has amplitude and phase place or be equal to a real part and an imaginary part.Therefore, multiple first sub-band signals 2303 that the input audio signal on line 2300 is produced by analysis filterbank 2302 represent.

The subset of all first sub-band signals is input in a synthesis filter banks 2304.Synthesis filter banks 2304 has Ms passage, and wherein Ms is less than M.Therefore, all sub-band signals not produced by bank of filters 2302 are transfused to synthesis filter banks 2304, but only have a subset, that is, the passage of indicated by 2305 one specific lesser amt.In Figure 23 embodiment, subset 2305 contains a specific middle bandwidth, but selectively, subset also can contain start from bank of filters 2302 filter bank channel 1 until a channel number is less than a bandwidth of the passage of M, or selectively, subset 2305 also can contain consistent with most high channel M and be extended the sub-band signal group of channel number higher than the low passage of logical number 1.Alternatively, passage index can from 0 according to the number scale in fact used.But preferably, for bandwidth expansion operation, the specific middle bandwidth that the sub-band signal group represented by 2305 represents is input in synthesis filter banks 2304.

Other passage not belonging to group 2305 does not input in synthesis filter banks 2304.Synthesis filter banks 2304 produces a middle sound signal 2306, and it has and equals f _sm _sthe sampling rate of/M.Due to M _sless than M, the sampling rate of input audio signal that the sampling rate of M signal 2306 will be less than on line 2300.Therefore, M signal 2306 represents the down-sampling corresponding with the bandwidth signal that sub-band 2305 represents and the signal of demodulation, and wherein signal is demodulated to base band, this is because the minimum passage of group 2305 is transfused to M _sthe passage 1 of synthesis filter banks, and the most high channel of block 2305 is transfused in the highest input of block 2304, and have nothing to do with the Aliasing Problem that some zero paddings for minimum or most high channel operate the border to avoid subset 2305.In addition, the device processing an input audio signal comprises another analysis filterbank 2307, and for analyzing M signal 2306, and this another analysis filterbank has M _aindividual passage, wherein M _abe different from M _sand be preferably more than M _s.Work as M _abe greater than M _stime, then this another analysis filterbank 2307 export with the sampling rate of 2308 sub-band signals represented by the sampling rate lower than a sub-band signal 2303.But, work as M _alower than M _stime, then the sampling rate of a sub-band signal 2308 is by the sampling rate of the sub-band signal higher than multiple first sub-band signal 2303.

Therefore, the cascade of bank of filters 2304 and 2307 (and preferably 2302) provides very efficient and high-quality up-sampling or down-sampling operation, or usually provides one resampling handling implement very efficiently.Multiple second sub-band signal 2308 is preferably further processed in a processor 2309, and the data of this processor to the cascade resampling of filtered device group 2304,2307 (and preferably 2302) perform process.In addition, preferably, block 2309 also performs the up-sampling operation being used for bandwidth expansion process operation, and the sub-band that the sub-band that last block 2309 is exported and block 2302 export has identical sampling rate.Then, in a bandwidth expansion process application, these sub-bands are transfused to a composite wave device group 2311 together with the other sub-band represented with 2310, this other sub-band is such as preferably the Low-frequency band frequency band produced by analysis filterbank 2302, composite wave device group 2311 finally provides a process time-domain signal, and such as one has a sampling rate 2f _sbandwidth expansion signal.This sampling rate of block 2311 output is the twice of the sampling rate of the signal on line 2300 in the present embodiment, and the sampling rate exported by block 2311 makes enough greatly the extra bandwidth produced by the process in block 2309 can represent with the process time-domain signal with high audio quality.

According to the application-specific of the present invention of cascading filter group, bank of filters 2302 can in a specific installation, and one only can comprise synthesis filter banks 2304 and another analysis filterbank 2307 for the treatment of the device of input audio signal.In other words, analysis filterbank 2302 and one " afterwards " processor can be separation distributions, should comprise block 2304,2307 by " afterwards " processor, and according to enforcement, also can comprise block 2309 and 2311.

In other embodiments, the difference implementing the application of the present invention of cascading filter group is, one specific device comprises analysis filterbank 2302 and less synthesis filter banks 2304, and M signal is provided to a different processor distributed by different distributor or is provided by a different passage that distributes.So the combination of analysis filterbank 2302 and less synthesis filter banks 2304 represents one down-sampling mode very efficiently, and the bandwidth signal solution representated by subset 2305 is transferred to base band simultaneously.This down-sampling and being performed to the demodulation of base band, and lose without any audio quality, and particularly without any audio-frequency information loss, be therefore a high-quality process.

Table in Figure 23 illustrates the particular exemplary number of different device.Preferably, analysis filterbank 2302 has 32 passages, and synthesis filter banks has 12 passages, and the number of active lanes that another analysis filterbank has is the twice of synthesis filter banks, such as 24 passages, and last synthesis filter banks 2311 has 64 passages.Generally speaking, the number of active lanes of analysis filterbank 2302 is comparatively large, and the number of active lanes of synthesis filter banks 2304 is less, and the number of active lanes of another analysis filterbank 2307 is placed in the middle and the number of active lanes of synthesis filter banks 2311 is very large.The sampling rate of the sub-band signal that analysis filterbank 2302 exports is f _s/ M.M signal has sample rate f _sm _s/ M.The sub-band passage of another analysis filterbank represented with 2308 has a sample rate f _sm _s/ (MM _a), and when the process in block 2309 makes sampling rate double, synthesis filter banks 2311 provides sampling rate to be 2f _soutput signal.But when sampling rate does not double by the process in block 2309, then the sampling rate exported by synthesis filter banks will be correspondingly lower.Subsequently, will discuss about present invention further optimization embodiment.

Figure 14 shows the principle of the transposition based on sub-band block.Input time-domain signal is fed to the analysis filterbank 1401 providing a large amount of complex value sub-band signal.These complex value sub-band signals are fed to sub-band processing unit 1402.This large amount of complex value exports sub-band and is fed to synthesis filter banks 1403, the time-domain signal of itself and then output modifications.The sub-band process that sub-band processing unit 1402 performs based on non-linear block operates, and is the transposition version of input signal corresponding to transposition rank T > 1 to make the time-domain signal revised.Idea based on the sub-band process of block once defines the nonlinear operation that the block of more than one subband samples carries out by comprising, and wherein follow-up block is by window and overlap-add exports sub-band signal to produce.

Bank of filters 1401 and 1403 can be any complex-exponential-modulation type, such as QMF or window DFT.They can be superposed by even number or odd number in modulation, and can be defined by the prototype filter of a wide region or window.Importantly know the quotient of following two filter-bank parameters measured with physical unit.

△ f _s: the sub-band frequency spacing of analysis filterbank 1401;

△ f _a: the sub-band frequency spacing of synthesis filter banks 1403.

For the configuration of sub-band process 1402, need to find out the corresponding relation between source and target sub-band index.Observe, the input sinusoidal curve of a physical frequencies Ω will cause having index m ≈ T Ω/△ f _sinput sub-band there is main contributions.Need the output sinusoidal curve of the physical frequencies T Ω of transposition will have index m ≈ T Ω/△ f by feeding _ssynthon frequency band produce.Therefore, the suitable source sub-band desired value of the sub-band process of a specific objective sub-band index m must be observed

$n\≈\frac{{\mathrm{\Δf}}_S}{{\mathrm{\Δf}}_A}\·\frac{1}{T}m---\left(1\right)$

Figure 15 shows an exemplary scenario of the application of the transposition based on sub-band block using the transposition of number rank in an enhancement mode HFR audio codec.One transmission bit stream is received by core decoder 1501, and this core decoder provides the core signal of a low bandwidth decoding with sample frequency fs.Low frequency by multiple modulation 32 frequency band QMF analysis bank 1502 resampling before one 64 frequency band QMF synthesis groups (reverse QMF) 1505 to output sampling frequency rate 2f _s.This two bank of filters 1502 and 1505 has identical physical resolution parameter △ f _s=△ f _a, and HFR processing unit 1504 only makes the unmodified lower sub-band corresponding to low bandwidth core signal pass through.The radio-frequency component of output signal by feed to the higher sub-band of 64 frequency band QMF synthesis groups 1505 from multiple transposition device unit 1503, the output band that performed spectrum shaping and amendment by HFR processing unit 1504 obtains.The core signal of decoding is regarded input and is exported the multiple sub-band signals representing the superposition of some transposition signal contents or the 64QMF frequency range analysis of combination by multiple transposition device 1503.Object is if HFR process is skipped over, then each composition is equivalent to an integer physics transposition of core signal, (T=2,3 ...).

Figure 16 shows the prior art exemplary scenario of the operation of the multistage transposition 1603 based on sub-band block, and an independent analysis filterbank is applied on each transposition rank of this operation.Herein, in the territory of the 64 frequency band QMF operated with 2fs sampling rate by generation three transposition rank T=2,3,4 and three transposition rank T=2,3,4 be output.Merge cells 1604 is only selected and is merged by the correlator frequency band from each transposition factor branch road to become one by by the single QMF group of subbands of HFR processing unit of feeding.

First consider the situation of T=2, particularly, object be one 64 frequency band QMF analyze 1602 ?2, one sub-band processing unit 1603 ?the 2 and 1 frequency band QMF processing chain of synthesizing 1505 produce the physics transposition of a T=2.Be 1401,1402 and 1403 by these three block identifications in Figure 14, find △ f _s/ △ f _a=2 make according to (1) cause 1603 ?2 the source that is specially n and target sub-band m between corresponding relation be given n=m.

As for the situation of T=3, example system comprise a sampling rate converter 1601 ?3, input sampling rate down coversion Graph One factor 3/2 makes to become 2fs/3 from fs by it.Particularly, object be this 64 frequency band QMF analyze 1602 ?3, this sub-band processing unit 1603 ?the 3 and 1 frequency band QMF processing chain of synthesizing 1505 cause the physics transposition of T=3.Be 1401,1402 and 1403 by these three block identifications in Figure 14, find due to resampling △ f _s/ △ f _a=3, make (1) provide 1603 ?3 the source that is specially n and target sub-band m between corresponding relation be again given n=m.

For the situation of T=4, example system comprise a sampling rate converter 1601 ?4, it, by input sampling rate down coversion Graph One factor 2, becomes fs/2 from fs.Particularly, object be this 64 frequency band QMF analyze 1602 ?4, this sub-band processing unit 1603 ?the 4 and 1 frequency band QMF processing chain of synthesizing 1505 cause the physics transposition of a T=4.Be 1401,1402 and 1403 by these three block identifications in Figure 14, find due to resampling △ f _s/ △ f _a=4, make (1) provide 1603 ?4 the source that is specially n and target sub-band m between corresponding relation be also given n=m.

Figure 17 shows the invention exemplary scenario of the efficient operation of the multistage transposition based on sub-band block of the single 64 frequency band QMF analysis filterbank of application one.In fact, use three independent QMF analysis bank and two sampling rate converters to cause a quite high computation complexity in Figure 16, and due to sample rate conversion 1601 ?3 cause cause some shortcomings implemented based on frame (frame) process.Current embodiment teach respectively with sub-band process 1703 ?3 and 1703 ?4 replace two branch roads 1601 ?3 → 1602 ?3 → 1603 ?3 and 1601 ?4 → 1602 ?4 → 1603 ?4, but branch road 1602 ?2 → 1603 ?2 to remain unchanged compared with Figure 16.Three all rank transposition must perform at present in a filter-bank domain with reference to Figure 14, wherein △ f _s/ △ f _a=2.With regard to the situation of T=3, provided by (1) 1703 ?3 specifically source n and target sub-band m between corresponding relation be given n ≈ 2m/3.With regard to the situation of T=4, provided by (1) 1703 ?4 specifically source n and target sub-band m between corresponding relation be given n ≈ 2m.In order to further reduce complexity, some transposition rank produce by the output copying transposition rank or the core decoder calculated.

Fig. 1 show a HFR enhanced decoder framework (such as SBR [ISO/IEC 14496 ?3:2009, Xin ceases the Bian Ma – Part III that Ji Shu – sound looks object: audio frequency]) in, use the operation of the transposition device based on sub-band block on 2,3 and 4 transposition rank.Bit stream is decoded to time domain by core decoder 101 and is sent to HFR module 103, and it produces a high-frequency signal by base band core signal.After generation, the signal that HFR produces is dynamically adjusted for mating original signal as closely as possible by the side information transmitted.By HFR processor 105, this adjustment is performed to the sub-band signal obtained from one or several analysis QMF group.A typical scheme is that wherein the time-domain signal of core decoder to the half frequency sampling with an input and output signal operates, that is, HFR decoder module will reach twice sample frequency by resampling core signal efficiently.The first step 102 that the conversion of this sample rate normally carries out filtering by one 32 frequency range analysis QMF groups, 102 pairs of core encoder signals obtains.Sub-band (that is, the lower subsets of 32 sub-bands containing whole core encoder signal energy) below so-called crossover frequency with carry the collective combinations that HFR produces the sub-band of signal.Usually, the sub-band number of so combination is 64, produces the core encoder signal of the sample rate conversion that combines with the output from HFR module after via the filtering of synthesis QMF group 106.

In HFR module 103 based in the transposition device of sub-band block, three transposition rank T=2,3 and 4 will produce and be transmitted in the territory of the 64 frequency band QMF operated with output sampling rate 2fs.Input time-domain signal block 103 ?12,103 ?13 and 103 ?in 14 by bandpass filtering.Carry out this action to make, by the output signal of different transposition rank process, be there is non-overlapped spectrum component.Signal by further down-sampling (103 ?23,103 ?24), the sampling rate of input signal to be adjusted to the analysis filterbank of an applicable fixed size (for 64 in this situation).Note, the increase of the sampling rate from fs to 2fs can use down-sampling factor T/2 by sampling rate converter but not the fact of T is explained, wherein the latter will produce the transposition sub-band signal with the sampling rate equal with input signal.Down-sampled signal by feed be separated HFR analysis filterbank (103 ?32,103 ?33 and 103 ?34), one for transposition rank, this bank of filters provides multiple complex value sub-band signal.These signals fed nonlinearities electric band spread unit (103 ?42,103 ?43 and 103 ?44).Multiple complex value output sub-band is fed and is merged/composite module 104 together with the output of sub sampling analysis bank 102.Sub-band from core analysis bank of filters 102 is only merged into one with each broadening factor branch road by merging/assembled unit will be fed to the single QMF group of subbands of one in HFR processing unit 105.

When the signal spectrum from different transposition rank is configured to not overlapping, that is, the frequency spectrum of T transposition rank signal should originate in T ?the frequency spectrum termination of 1 rank signal, the signal demand of transposition has bandpass characteristics.Conventional band-pass filters in Fig. 1 103 ?12 ?103 ?14 to come therefrom.But simply get rid of selection via merging/assembled unit 104 can utilize in sub-band one, independent bandpass filter is unnecessary and can be removed.Alternatively, the intrinsic bandpass characteristics provided by QMF group is by feeding the different sub-band in 104 and being utilized independently by the contribution of the difference of transposition device branch road.Only the band applications time explanation be combined in 104 is also satisfied the demands.

Fig. 2 shows the operation of a nonlinearities electric band spread unit.Block extraction apparatus 201 is sampled from complex value input signal a limited frame of a sample.Frame is defined by an input pointer position.This frame accept Nonlinear Processing in 202 and then in 203 by finite length window window.The sample produced is added into previous output sample in overlap and adder unit 204, and wherein output frame position is defined by an output pointer position.Input pointer increases with a fixed amount and output pointer is multiplied by the increase of identical amount with this sub-band broadening factor.The repeating of this operational chain will cause a duration to be the output signal that sub-band broadening factor is multiplied by the input sub-band signal time, and duration of output signal is up to the length of synthesis window.

Although SBR [ISO/IEC 14496 ?3:2009, Xin ceases the Bian Ma – Part III of Ji Shu – sound depending on object: audio frequency] the SSB transposition device that uses typically utilizes whole base band except the first sub-band to produce high-frequency band signals, but harmonic wave transposition device uses smaller portions of core encoder frequency spectrum usually.The amount (so-called source range) used depends on transposition rank, bandwidth expansion factor and is applicable to the rule of combined result, such as, whether allow the signal spectrum overlap produced by different transposition rank.Therefore, in fact harmonic wave transposition device will be used by HFR processing module 105 with regard to an only finite part of the output spectrum on specific transposition rank.

Figure 18 shows another embodiment that the exemplary process for the treatment of single sub-band signal is implemented.Single sub-band signal, before or after be not shown in the analysis filterbank filtering in Figure 18 by one, receives the extraction of any type.Therefore, the time span of single sub-band signal is shorter than forming the time span before extracting.Single sub-band signal is input in a block extraction apparatus 1800, and this extraction apparatus can be identical with block withdrawal device 201, but also can implement by different way.Block extraction apparatus 1800 in Figure 18 uses one to be exemplarily called, and the sample/block prior value of e operates.This sample/block prior value can be variable or can be fixing setting, and shown in Figure 18 be arrow in a sensing block extraction apparatus block 1800.In the output of block extraction apparatus 1800, there is multiple extraction block.These blocks are overlapping to heavens, this is because sample/block prior value e is less than the block length of block extraction apparatus significantly.One example is the block that block extraction apparatus extracts 12 samples.First block comprise sample 0 ?the 11, second block comprise sample 1 ?the 12, three block comprise sample 2 ?13, etc.In this embodiment, sample/block prior value e equals 1, and has one 11 heavy overlaps.

Each block is transfused to window device 1802, to use a window function to make block window for each block.In addition, arrange phase calculator 1804, it calculates a phase place of each block.Phase calculator 1804 can use each block before window or after window.Then, phase adjustment value p x k is calculated and is transfused in a phase regulator 1806.Adjusted value is applied to each sample in block by phase regulator.In addition, factor k equals bandwidth expansion factor.For example, when to obtain Graph One factor be the bandwidth expansion of 2, then the phase place p that the block extracted for block extraction apparatus 1800 calculates is multiplied by 2 and the adjusted value being applied to each sample of block in phase regulator 1806 is that p is multiplied by 2.This is an example value/rule.Or the correction phase place of synthesis is k*p, p+ (k ?1) * p.Therefore in this example, if taken advantage of calculation, correction factor is 2, if added, is then 1*p.Other value/rule can be applied to calculate phase correcting value.

In one embodiment, single sub-band signal is a multiple sub-band signal, and the phase place of a block can calculate in a number of different ways.Method adopts in the middle of block or the middle sample around of block, and calculate the phase place of the plurality of sample.Phase place can also be calculated for each sample.

Although figure 18 illustrates a phase regulator to operate after window device, these two pieces also can be exchanged, and make the onblock executing phase place adjustment of extracting block extraction apparatus, and then perform windowization operation.Due to two operations, namely window and phase place adjustment are real-valued or complex value multiplication algorithm, and these two operations are by using a Complex Multiplication Algorithm factor to be generalized into a single operation, and this Complex Multiplication Algorithm factor itself is the product of the phase place adjustment multiplication algorithm factor and a window factor.

Phase place adjustment block is transfused to one overlap/addition and correction of amplitude block 1808, wherein the superimposed addition of block of this window and adjustment phase place.But the more important thing is, the sample in block 1808/block prior value is different from the value used in block extraction apparatus 1800.Especially, the sample in block 1808/block prior value is greater than the value e used in block 1800, so obtain the time explanation of the signal that block 1808 exports.Therefore, the length of the processed sub-band signal of block 1808 output is longer than the length of the sub-band signal inputed in block 1800.When obtaining one and be the bandwidth expansion of two, then use sample/block prior value, this prior value is the twice of the respective value in block 1800.This cause Graph One factor be two time explanation.But, when needs At All Other Times broadening factor time, then can use other sample/block prior value, to make the output of block 1808, there is required time span.

In order to solve overlap problem, preferably perform correction of amplitude, to solve the problem of the not negative lap in block 1800 and 1808.But this correction of amplitude also can be introduced in window device/phase regulator multiplication algorithm factor, but correction of amplitude also can perform after overlapping/process.

Be 12 and sample/block prior value in block extraction apparatus is in the example of in an above-mentioned block length, when to perform the factor be the bandwidth expansion of 2, the sample/block prior value of overlap/addition block 1808 will equal two.This will cause the overlap of five blocks.When being the bandwidth expansion of 3 by the execution factor, then sample/block prior value that block 1808 uses will equal three, and overlap will drop to 3 overlaps.When by execution four times of bandwidth expansions, then overlap/addition block 1808 must use be four sample/block prior value, it will cause the overlap more than two blocks.

Be constrained to by the input signal of near transposition device branch road and only comprise source range and can realize a large amount of calculated savings, and this is adapted to each transposition rank under a sampling rate.This system for one based on the HFR generator of sub-band block fundamental block design shown in Figure 3.Input core coded signal is by the special down-sampler process before HFR analysis filterbank.

The Essential Action of each down-sampler is filtering source range signal, and by it so that minimum sampling rate analysis filterbank may be sent to.Herein, " may be minimum " refers to the Least sampling rate being still suitable for downstream, needs not to be the Least sampling rate avoiding the aliasing after extracting.Sample rate conversion can obtain in every way.Under the prerequisite not limiting the scope of the invention, two examples will be provided: the first example provides the resampling undertaken by multi tate Time Domain Processing, and the second example illustrates the resampling realized by the process of QMF sub-band.

Fig. 4 shows the example that transposition rank are the block in the multi tate time domain down-sampler of two.There is bandwidth B hertz and sample frequency be fs input signal by a complex exponential (401) modulation, to make the beginning frequency displacement of source range to DC frequency, as follows:

$x_m\left(n\right)=x\left(n\right)\·\exp (-i2\mathrm{\π}{f}_s\frac{B}{2})$

An input signal after modulation and the example of frequency spectrum are shown in Fig. 5 (a) and (b).Modulation signal is by interpolation (402) and by a complex value low-pass filter to be with logical restriction 0 and B/2 hertz filtering (403).Frequency spectrum after each step is illustrated in Fig. 5 (c) and (d).Filtering signal is then extracted (404), and the real part of signal is calculated (405).The results are shown in Fig. 5 (e) and (f) figure after these steps.In this special example, work as T=2, during B=0.6 (in a normalization scale, namely fs=2), in order to safety contains source range, P ₂be selected as 24.The down-sampling factor obtains:

$\frac{32T}{P_2}=\frac{64}{24}=\frac{8}{3}$

, its mid-score has used common factor 8 abbreviation, and therefore, interpolation factor was 3 (as Fig. 5 (c) is visible), and the extraction factor is 8.By using Noble identical relation [" Multirate Systems And FilterBanks, " P.P.Vaidyanathan, 1993, Prentice Hall, Englewood Cliffs], withdrawal device can be moved to left always in the diagram, and interpolator can be moved to right always.So, may minimum sampling rate carry out modulating and filtering, and computation complexity is further reduced.

Another approach is that the sub-band being used in already present sub sampling 32 frequency range analysis QMF group 102 in SBR HFR method exports.The sub-band containing the source range of different transposition device branch road is combined into time domain by the QMF of the little sub sampling before HFR analysis filterbank.This HFR system is shown in Figure 6.Little QMF group is obtained by the 64 frequency band QMF groups that sub sampling is original, and wherein prototype filter coefficient is found out by the linear interpolation method of original prototype filter.Note the symbol in Fig. 6, the synthesis QMF group before second-order transposition device branch road has Q ₂=12 frequency bands (there is the sub-band of zero-base index 8 to 19 in 32 frequency band QMF).Aliasing in synthesis process, first (index 8) and last (index 19) frequency band are set as zero.The frequency spectrum produced exports and illustrates at Fig. 7.Note, the transposition device analysis filterbank based on block has 2Q ₂=24 frequency bands are that is, identical with the number of the frequency band in the example (Fig. 3) based on multi tate time domain down-sampler.

As comparison diagram 6 and Figure 23, be apparent that, the element 601 of Fig. 6 corresponds to the analysis filterbank 2302 of Figure 23.In addition, the synthesis filter banks of Figure 23 2304 correspond to element 602 ?2, and another analysis filterbank 2307 of Figure 23 correspond to element 603 ?2.Block 604 ?2 correspond to block 2309, and combiner 605 may correspond in synthesis filter banks 2311, but in other embodiments, combiner can be configured to export sub-band signal, and then, can use another synthesis filter banks being connected to combiner.But, according to enforcement, discuss in the background of Figure 26 after a while one specific high-frequency reconstruction can perform before synthesis filter banks 2311 carries out synthetic filtering or combiner 205, or can perform after the combiner after bank of filters 2311 synthetic filtering of Figure 23 or in the block 605 of Fig. 6.

From 602 ?3 extend to 604 ?3 or from 602 ?T extend to 604 ?other branch road of T not shown in fig 23, but to implement in a similar manner, but the varying in size of the bank of filters had, the T-phase wherein in Fig. 6 is when in a transposition factor.But, as in the background of Figure 27 discuss, the transposition that the transposition factor is 3 and the transposition that the transposition factor is 4 can be introduced into by element 602 ?2 to 604 ?in the 2 process branch roads formed, therefore block 604 ?2 not only provide Graph One factor be 2 transposition, and provide together with the specific synthesis filter banks used discussed in the background of 27 at Figure 26 Graph One factor be 3 and Graph One factor be the transposition of 4.

In Fig. 6 embodiment, Q ₂corresponding to Ms, and Ms equals such as 12.In addition, another analysis filterbank 603 ?2 size correspond to element 2307, equaling 2Ms, such as, is 24 in the present embodiment.

In addition, as front general introduction, lowest subband passage and the highest sub-band passage of synthesis filter banks 2304 can be fed with zero, to avoid Aliasing Problem.

The system outlined in FIG can be regarded as one of the resampling summarized in Fig. 3 and Fig. 4 and simplify special case.In order to simplify configuration, omit modulator.Further, the analysis filterbank of 64 frequency bands is used to obtain all HFR analysis filtered.Therefore, the P of Fig. 3 ₂=P ₃=P ₄=64, and second, third is 1,1.5 and 2 with the down-sampling factor of the 4th rank transposition device branch road respectively.

An advantage of the present invention is under the threshold sampling processing environment of invention, from the block 2302 corresponded in Figure 23 or as MPEG ?in 4 (ISO/IEC14496 ?3) the sub-band signal of 32 frequency range analysis QMF groups of 601 in Fig. 6 of defining can be used.The definition of this analysis filterbank in MPEG ?4 illustrates on the top of Figure 25 a, and is shown as a process flow diagram of Figure 25 b, and it also takes MPEG ?4 standard.SBR (spectral bandwidth copies) part of this standard is incorporated herein by quoting as proof.Particularly, the analysis filterbank 2302 of Figure 23 or the 32 frequency band QMF 601 of Fig. 6 can be embodied as the process flow diagram of Figure 25 a top, Figure 25 b.

In addition, the synthesis filter banks shown in block 2311 of Figure 23 also can be implemented as shown in the bottom of Figure 25 a and as shown in the process flow diagram of Figure 25 c.But, can apply any other bank of filters definition, but at least with regard to analysis filterbank 2302, being embodied as illustrated by Figure 25 a and 25b is preferred, this is because at least bandwidth expansion application environment (such as spectral bandwidth copies, or typically, reconstruction process apply) in, have 32 passages MPEG ?4 analysis filterbank robustness, stability and high-quality are provided.

Synthesis filter banks 2304 is configured to the subset synthesis of the sub-band of the source range by containing transposition device.Carrying out this synthesis is in order to time history synthesis M signal 2306.Preferably, synthesis filter banks 2304 is the real-valued QMF group of a little sub sampling.

The time domain of this bank of filters exports 2306 complex values being then fed to bank of filters twice size and analyzes QMF group.This QMF group is illustrated by the block 2307 of Figure 23.This can realize the significantly saving of computation complexity when only having relevant source range to be switched to and having the QMF sub-band domain of double frequency resolution.Obtain little QMF group by the sub sampling of the QMF group of original 64 frequency bands, wherein prototype filter coefficient is obtained by the linear interpolation of original prototype filter.Preferably, with have 640 samples MPEG ?the prototype filter that associates of 4 synthesis filter banks used, wherein MPEG ?4 analysis filterbank there is the window of 320 window samples.

The process of descriptor sampling filter group in Figure 24 a that process flow diagram is shown and Figure 24 b.First following variable is determined:

M _S＝4·floor{(f _TableLow(0)+4)8+1}

k _L＝startSubband2kL(f _TableLow(0))

Wherein, MS is the size of sub sampling synthesis filter banks, and kL represents that the index of the first passage of 32 frequency band QMF groups is to enter sub sampling synthesis filter banks.Array startSubband2kL lists in table 1.Function f loor{x} round up independent variable x be towards negative infinitely great direction closest to integer.

Table 1 ?y=startSubband 2kL (x)

x	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30	31
																																	y	0	0	0	0	0	0	0	2	2	2	4	4	4	4	4	6	6	6	8	8	8	8	8	10	10	10	12	12	12	12	12	12

Therefore, M is worth _sthe size of the synthesis filter banks 2304 of definition Figure 23, and K _lit is the first passage of the subset 2305 represented in Figure 23.Clearly, the value in equation ftableLow be defined in ISO/IEC 14496 ?3, in part 4.6.18.3.2, incorporated herein by quoting as proof this part.Note, value M _sincrease progressively with 4, the size meaning synthesis filter banks 2304 can be 4,8,12,16,20,24,28 or 32.

Preferably, synthesis filter banks 2304 is real-valued synthesis filter banks.For this reason, according to the first step of Figure 24 a from M _snew complex value subband samples calculates M _sthe set of real-valued subband samples.For this reason, following equation is used:

$V(k-k_L)=\mathrm{Re}\{X_{\mathrm{Low}}\left(k\right)\&CenterDot;\exp \left(i\frac{\mathrm{\&pi;}}{2}(k_L-\frac{(k+0.5)\&CenterDot;191}{64})\right)\},k_L\&le;k<k_L+M_S$

In equation, exp () indicates complex-exponential function, and i is imaginary unit, and k _las defined above.

The sample in array v is made to move 2M _sposition.The oldest 2M _sindividual sample is dropped.

The real-valued subband samples of MS is multiplied with matrix N, that is, Ju Zhen ?vector product NV calculated, wherein

$N(k,n)=\frac{1}{M_S}\&CenterDot;\cos \left(\frac{\mathrm{\&pi;}\&CenterDot;(k+0.5)\&CenterDot;(2\&CenterDot;n-M_S)}{2M_S}\right),\left\{\begin{array}{c}0\&le;k<M_S\\ 0\&le;n<2M_S\end{array}\right.$

The output of this computing is stored in position 0 to the 2M of array v _s?1.

Sample is extracted to produce 10M from v according to the process flow diagram in Figure 24 a _sthe array g of individual element.

The sample of array g is multiplied by window ci to produce array w.Window coefficient ci is obtained by the linear interpolation (namely by following equation) of coefficient c:

c _i(n)＝ρ(n)c(μ(n)+1)+(1-ρ(n))c(μ(n)),0≤n<10M _S

Wherein μ (n) and ρ (n) is defined by 64n/M respectively _sinteger and fractional part.Window coefficient c can ISO/IEC 14496 ?3:2009 table 4.A87 in find.

Therefore, synthesis filter banks has a prototype window function counter, for carrying out sub sampling for the memory window function of the bank of filters with different size or interpolation calculates a prototype window function by using.

The sample read group total M to array w is passed through according to the final step of the process flow diagram in Figure 24 a _snew output sample.

Then, the preferred enforcement of another analysis filterbank 2307 in Figure 23 is shown in Figure 24 b together with process flow diagram.

The sample in array x is made to move 2M according to the first step of Figure 24 b _sindividual position.The oldest 2M _sindividual sample is dropped, and 2M _sindividual new samples is stored in position 0 to 2M _sin-1.

The sample of array x is multiplied by window coefficient c _2i.Window coefficient c _2iby the linear interpolation of coefficient c, that is obtained by following equation:

c _2i(n)＝ρ(n)c(μ(n)+1)+(1-ρ(n))c(μ(n)),0≤n<20M _S

Wherein μ (n) and ρ (n) is defined by 32n/M respectively _sinteger and fractional part.Window coefficient c can ISO/IEC 14496 ?3:2009 table 4.A87 in find.

Therefore, another analysis filterbank 2307 has a prototype window function counter, for by using one to carry out sub sampling for the memory window function with the bank of filters of different size or interpolation calculates a prototype window function.

Formula according to the process flow diagram in Figure 24 b is sued for peace to sample, to produce 4M _sthe array u of individual element.

By Ju Zhen ?multiplication of vectors Mu calculate 2M _sindividual new complex value subband samples, wherein

$M(k,n)=\exp \left(\frac{i\&CenterDot;\mathrm{\&pi;}\&CenterDot;(k+0.5)\&CenterDot;(2\&CenterDot;n-{4\&CenterDot;M}_S)}{4M_S}\right),\left\{\begin{array}{c}0\&le;k<{2M}_S\\ 0\&le;n<4M_S\end{array}\right.$

In equation, exp () represents complex-exponential function, and i is imaginary unit.

In Fig. 8 (a), illustrated that the factor is the block figure of the down-sampler of 2.New real-valued low-pass filter can be written to H (z)=B (z)/A (z), and wherein B (z) is non-recursive part (FIR) and A (z) is recursive component (IIR).But in order to efficient enforcement, use Noble identical relation to reduce computation complexity, design wherein all limits has tuple 2 (duopole) (as A (z ²)) wave filter be useful.Therefore wave filter can be broken down into shown in Fig. 8 (b).Use Noble identical relation 1, recursive component can be moved through withdrawal device, as in Fig. 8 (c).Nonrecursive filter B (z) can use 2 composition poly phase of standard to be implemented as:

$B\left(z\right)=\underset{n=0}{\overset{N_z}{\mathrm{\&Sigma;}}}b\left(n\right)z^{-n}=\underset{l=0}{\overset{5}{\mathrm{\&Sigma;}}}{z}^{-l}{E}_l\left(z^6\right),$ Wherein $E_l\left(z\right)=\underset{n=0}{\overset{N_z/6}{\mathrm{\&Sigma;}}}b(6\&CenterDot;n+l)z^{-n}$

Therefore, down-sampler can be configured to as shown in Fig. 8 (d).After use Noble identical relation 1, so that FIR part may be calculated, as shown in Fig. 8 (e) by Least sampling rate.Can easily find out from Fig. 8 (e), FIR operation (postpone, extract and multi-phase components) can be regarded as the window ?phase add operation of a use two sample input step-length.For two input amendment, a new output sample will be generated, with the down-sampling of realization factor 2 efficiently.

One piece of factor 1.5=3/2 down-sampler illustrates in Fig. 9 (a).Real-valued low-pass filter can be written to H (z)=B (z)/A (z) once again, and wherein B (z) is non-recursive part (FIR) and A (z) is recursive component (IIR).As aforementioned, in order to efficient enforcement, use Noble identical relation to reduce computation complexity, design wherein all limits or there is tuple 2 (duopole) or tuple 3 (three limits) (as A (z ²) or A (z ³)) a wave filter be useful.Herein, the algorithm for design that duopole is selected as low-pass filter is more efficient, but three limit modes are compared, and recursive component has 1.5 times of complexities in fact on the implementation.Therefore wave filter can be broken down into as shown in Fig. 9 (b).Use Noble identical relation 2, recursive component can move before interpolator, as shown in Fig. 9 (c).Nonrecursive filter B (z) can use standard 23=6 composition poly phase to be implemented as:

$B\left(z\right)=\underset{n=0}{\overset{N_z}{\mathrm{\&Sigma;}}}b\left(n\right)z^{-n}=\underset{l=0}{\overset{5}{\mathrm{\&Sigma;}}}{z}^{-l}{E}_l\left(z^6\right),$ Wherein $E_l\left(z\right)=\underset{n=0}{\overset{N_z/6}{\mathrm{\&Sigma;}}}b(6\&CenterDot;n+l)z^{-n}$

Therefore, down-sampler can be configured to as shown in Fig. 9 (d).After use Noble identical relation 1 and 2, so that FIR part may be calculated, as shown in Fig. 9 (e) by Least sampling rate.Easily find out from Fig. 9 (e), use three multiphase filter E of lower group ₀(z), E ₂(z), E ₄z () calculates even number index output sample, and higher group of E ₁(z), E ₃(z), E ₅z () calculates odd number indexed samples.Often the operation of group (delay chain, withdrawal device and heterogeneous element) can be regarded as the window ?phase add operation of the input step-length of use three sample.The window coefficient that upper set uses is odd number index coefficient, and below group uses the odd number index coefficient from original filter B (z).Therefore, for the group of one or three input amendment, generation two new output samples will be generated, cause the down-sampling of the factor 1.5 efficiently.

Time-domain signal from core decoder (101 in Fig. 1) is also converted by sub sampling by the sub sampling synthesis that use one in core decoder is less.A less synthesis conversion is used to provide the reduction of further computation complexity.According to crossover frequency (i.e. the bandwidth of core encoder signal), the core encoder causing generation one to have sampling rate Qfs outputs signal by the ratio of synthesis change of scale and nominal dimension Q (Q<1).In the example summarized in this application, in order to process sub sampling core encoder signal, all analysis filterbank 1 (102 in Fig. 1,103 ?32,103 ?33 and 103 ?34), together with Fig. 3 down-sampler (301 ?2,301 ?3 and 301 ?T), the withdrawal device 404 of Fig. 4 and the analysis filterbank 601 of Fig. 6 need with factor Q proportional zoom.Apparently, Q must be selected to all bank of filters sizes is integer.

Figure 10 show the spectral boundaries of HFR transposition signal and a HFR enhanced decoder (such as SBR [ISO/IEC 14496 ?3:2009, Xin cease the Bian Ma – Part III of Ji Shu – sound depending on object: audio frequency]) in envelope to adjust frequency the aligning shown.Figure 10 (a) shows the format chart of the frequency band comprising envelope adjustment form, and alleged scale factor contains from crossover frequency kx to the frequency range stopping frequency ks.Scale factor is formed in the frequency graticule mesh (frequency envelope) used when adjusting the energy level of regenerate high-frequency band frequency in a HFR enhancement mode scrambler.In order to adjust envelope, the signal energy in the time/frequency block limited by scale factor border and selected time boundary is averaged.If the signal misalignment scale factor produced by different transposition rank, as as shown in Figure 10 (b), due to the spectrum structure that envelope adjustment process will keep in scale factor, if so spectrum energy sharply changes near transposition frequency band border, then can artifact be caused.Therefore, the solution proposed is the border making the frequency boundary of transposition signal adapt to the scale factor shown in Figure 10 (c).In this figure, existing scale factor border lower than one in a small amount, is aimed to make the frequency boundary of transposition frequency band in the coboundary of the signal produced by transposition rank 2 and 3 (T=2,3) compared with Figure 10 (b).

The practical situation showing potential artifact when using non-aligned border has been shown in Figure 11.Figure 11 (a) also show scale factor border.Figure 11 (b) shows transposition rank T=2, the signal not adjusting HFR generation of 3 and 4 and core codec baseband signal.Figure 11 (c) shows the envelope adjustment signal when the smooth target envelope of employing one.The block with reticulate pattern region represents the scale factor with the change of high frequency band self-energy, and it can cause the exception of output signal.

Figure 12 shows the situation of Figure 11, but this time uses the border aimed at.Figure 12 (a) shows scale factor border, Figure 12 (b) shows transposition rank T=2, the signal not adjusting HFR generation of 3 and 4 and core codec baseband signal, and consistent with Figure 11 (c), Figure 12 (c) shows the envelope adjustment signal when the smooth target envelope of employing one.From this figure, because of the misalignment of transposition signal band and scale factor, cause there is not the scale factor with high frequency band self-energy, and therefore potential artifact is reduced.

Figure 13 show HFR limiter frequency band border (as, such as [SBR [and ISO/IEC 14496 ?3:2009, Xin cease the Bian Ma – Part III that Ji Shu – sound looks object: audio frequency] in describe] adaptation that the harmonic wave in a HFR enhancement mode scrambler is repaired.Limiter is to the frequency band operation with the resolution being far coarser than scale factor, but principle of operation is very identical.In limiter, the average gain value of each limiter frequency band is calculated.Each yield value (that is, for the envelope gain value that each scale factor calculates) does not allow specific the taking advantage of exceeding limiter average gain value to calculate more than the factor.The object of limiter is the large change of the scale factor gain suppressed in each limiter frequency band.Although the adaptation that transposition device produces frequency band comparative example factor band guarantees that the inband energy change in a scale factor is little, but according to the present invention, the adaptation on limiter band edge bound pair transposition device frequency band border solves the comparatively large scale energy difference through between the frequency band of transposition device process.Figure 13 (a) shows transposition rank T=2, and the HFR of 3 and 4 produces the frequency limitation of signal.The energy water adjustment of different transposition signal can be different in essence.Figure 13 (b) shows the frequency band of limiter, and this limiter has fixed width about a logarithmic frequency scale typically.Transposition device frequency band border is added as fixing limiter border, and remaining limiter border is recalculated that logarithmic relationship is kept close as far as possible, as shown in the example of Figure 13 (c).Although describe in some under the background of device, obviously, these aspects also represent the description of corresponding method, and wherein one piece or device correspond to the feature of a method step or a method step.Similarly, the corresponding blocks of corresponding intrument or the description of item or feature is also represented in describing under the background of method step.

Other embodiment uses the mixing patch system shown in Figure 21, wherein performs the mixing method for repairing and mending in a time block.In order to contain the zones of different of HF frequency spectrum completely, BWE comprises several repairing.In HBE, higher repairing needs the high transposition factor in phase vocoder, and this reduces instantaneous perceptual quality especially.

Therefore, embodiment copies the higher-order repairing of repairing and producing and occupying top spectral regions preferably by the upper efficient SSB of calculating, and the lower-order repairing producing and contain intermediate spectral region is repaired preferably by HBE, wherein for intermediate spectral region, expect to keep harmonic structure.Indivedual mixing of method for repairing and mending in time through being static, or preferably can obtain signal in the bitstream.

About replicate run, low-frequency information can be used, as shown in figure 21.Or, can be used as shown in Figure 21 from the data of the repairing using HBE method to produce.The latter causes the more not intensive pitch structure for higher repairing.Except these two examples, it is all imaginabale for copying with often kind of combination of HBE.

Concept proposed advantage be

Improve instantaneous perceptual quality

Reduce computation complexity

Figure 26 shows the preferred process chain for bandwidth expansion, can perform different process operations in the nonlinearities frequency band process wherein represented at block 1020a, 1020b.In fig. 26, cascading filter group 2302,2304,2307 represents with block 1010.In addition, block 2309 may correspond in element 1020a, 1020b, and envelope adjuster 1030 can be arranged between the block 2309 of Figure 23 and block 2311, after maybe can being arranged on the process of block 2311.In this implements, the band selective process of the time-domain signal (such as bandwidth expansion signal) of process is in time domain but not perform in sub-band domain, before this sub-band domain is present in synthesis filter banks 2311.

Figure 26 shows the device producing bandwidth extended audio signal from a low band signal 1000 according to another embodiment.Device comprises analysis filterbank 1010, sub-band wise nonlinearities band processors 1020a, 1020b, one with latter linked envelope adjuster 1030 or, just generally speaking, high-frequency reconstruction parameter (such as, inputting on parameter line 1040) is carried out to the high-frequency reconstruction processor operated.Envelope adjuster, or just generally speaking, each sub-band signal of each sub-band of high-frequency reconstruction processor process, and will be used in the process sub-band signal input synthesis filter banks 1050 of each sub-band passage.Synthesis filter banks 1050 is at its lower channel reception input signal, and the sub-band of low-frequency band core decoder signal represents.According to enforcement, low-frequency band can also derive from the output of the analysis filterbank 1010 Figure 26.Transposition sub-band signal is fed to the comparatively high filter group passage of synthesis filter banks, to perform high-frequency reconstruction.

Bank of filters 1050 finally exports a transposition device output signal, it comprises the transposition factor 2, the bandwidth expansion of 3 and 4, and the signal that block 1050 exports no longer is limited to crossover frequency by bandwidth, be namely no longer restricted to the highest frequency of the core encoder signal of the low-limit frequency being equivalent to the signal content that SBR or HFR produces.

In Figure 26 embodiment, analysis filterbank performs twice with up-sampling, and has a specific analysis sub-band spacing 1060.Synthesis filter banks 1050 has a synthon frequency band spacing 1070, and in present embodiment, this makes analysis sub-band spacing size doubles, and this will cause the transposition contribution will discussed in the background of Figure 27 after a while.

Figure 27 shows the detailed enforcement of the preferred implementation of the nonlinearities band processors 1020a in Figure 26.Circuit shown in Figure 27 receives single sub-band signal 108 as an input, and this single sub-band signal 108 will be processed in three " branch road ": upper branch road 110a is used to the transposition factor 2 transposition.Be positioned at the branch road represented with 110b in the middle of Figure 27 to be used for the transposition factor 3 transposition, and the lower branch road represented with reference number 110c in Figure 27 is for the transposition factor 4 transposition.But the actual transposition obtained by each treatment element in Figure 27 is only 1 (namely not having transposition) to branch road 110a.By the treatment element shown in Figure 27 1.5 are equaled for the actual transposition that middle branch 110b obtains and 2 are equaled to the actual transposition that lower branch road 110c obtains.This is to be arranged in the numeral of Figure 27 left side bracket, wherein represents transposition factor T.1.5 and 2 transposition represent by branch road 110b, carry out extraction operation and to be carried out the first transposition that time explanation obtains with adder processor contributed by overlapping in 110c.Second contribution (i.e. transposition double) is obtained by synthesis filter banks 105, and this synthesis filter banks 105 has the synthon frequency band spacing 107 that doubles analysis filterbank sub-band spacing.Therefore, because synthesis filter banks has twice analyze sub-band spacing, so there is not any extract function in branch road 110a.

But branch road 110b has an extract function with the transposition obtaining 1.Because synthesis filter banks has the physics sub-band spacing doubling analysis filterbank, one obtains the transposition factor 3, as indicated the left of the block extraction apparatus of the second branch road 110b in figure 27.

Similarly, the 3rd branch road has one to correspond to the extract function of the transposition factor 2, and the final contribution of analysis filterbank and the different sub-band spacing in synthesis filter banks is corresponding to the transposition factor 4 of the 3rd branch road 110c.

Especially, each branch road has block extraction apparatus 120a, 120b, a 120c, and each in these block extraction apparatuss can be similar with the block extraction apparatus 1800 of Figure 18.In addition, each branch road has phase calculator 122a, 122b and a 122c, and phase calculator can be similar with the phase calculator 1804 of Figure 18.Moreover each branch road has phase regulator 124a, 124b, a 124c, and phase regulator can be similar with the phase regulator 1806 of Figure 18.In addition, each branch road has window device 126a, 126b, a 126c, and each of wherein these window devices can be similar with the window device 1802 of Figure 18.But window device 126a, 126b, 126c also can be configured to be applied to a rectangular window together with some " zero paddings ".In the embodiment of Figure 27, transposition in each branch road 110a, 110b, 110c or repair signal are transfused to totalizer 128, contribution from each branch road is added to current sub-band signal by totalizer 128, obtains so-called transposition block with the final output in totalizer 128.Then, Chong Die ?perform in totalizer 130 a Chong Die ?be added process, and Chong Die ?totalizer 130 can with Figure 18 overlapping/addition block 1808 is similar.Chong Die ?totalizer apply an overlap-add prior value 2e, wherein e be block extraction apparatus 120a, 120b, 120c Chong Die ?prior value or " step value ", and overlapping totalizer 130 exports the signal of transposition, it is the single sub-band output of a passage k (that is, the current sub-band passage observed) in the embodiment of Figure 27.Sub-band is analyzed for each or the process shown in Figure 27 is performed for a particular analysis group of subbands, and as shown in Figure 26, the sub-band signal of transposition is imported into synthesis filter banks 1050 after being processed by block 1030, and finally obtain transposition device output signal in the output of block 1050, as shown in Figure 26.

In embodiments, the block extraction apparatus 120a of the first transposition device branch road 110a extracts 10 subband samples, and subsequently, performs these 10 QMF subband samples to polar conversion.This output produced by phase regulator 124a is then sent to window device 126a, and window device 126a exports with being finally worth to expand with zero for the first value of block, and wherein, this operation is equal to (synthesis) window of the rectangular window of a length 10.Block extraction apparatus 120a in branch road 110a does not perform extraction.Therefore, the same sample spacing that the sample extracted by block extraction apparatus is extracted with them is mapped to extracted block.

But this is different for branch road 110b and 110c.Block extraction apparatus 120b preferably extracts the block of one 8 subband samples, and these 8 subband samples extracted in block is distributed with different subband samples spacing.Obtained by method of interpolation and extract the non-integer subband samples item of block, and the QMF sample so obtained is switched to polar coordinates together with the sample of interpolation and by phase regulator process.Then, perform the window in window device 126b once again, with the block expansion zero exported by phase regulator 124b for initial two samples and last two samples, this operation is equal to (synthesis) window of the rectangular window of a length 8.

Block extraction apparatus 120c be configured for extraction one have 6 subband samples time width and perform one extract the factor be the extraction of 2, perform QMF sample to polar conversion, and the operation once again in excute phase adjuster 124b, and export again with zero expansion, but be for a most junior three sub-frequency band sample and last three subband samples at present.This operation is equivalent to (synthesis) window of the rectangular window of a length 6.

The transposition of each branch road exports and is then added, to form the combination QMF exported by totalizer 128, and combination QMF output finally uses overlapping totalizer to be applied in block 130, wherein this overlap-add prior value or step value are the twice of the step value of previously described block extraction apparatus 120a, 120b, 120c.

Embodiment comprises by using based on the method for the harmonic wave transposition decoded audio signal of sub-band block, the method comprise by a M ?frequency range analysis bank of filters filtering is carried out to a core codec signal, to obtain the set of sub-band signal; Synthesized by the subset of sub sampling composite filter to described sub-band signal with a minimizing sub-band number, to obtain sub sampling source range signal.

One embodiment relates to a kind of method of aiming at the spectral boundaries utilized in parameter processing for spectral band border HFR being produced signal.

One embodiment relates to a kind of spectral band border for HFR being produced signal and envelope and to adjust frequency the method that the spectral boundaries shown aims at, and the method comprises: the HFR that search is no more than transposition factor T in envelope adjusts frequency table produces the highest border of the primary bandwidth restriction of signal; And use the highest border found to produce the frequency limitation of signal as the HFR of transposition factor T.

One embodiment relates to a kind of method for the spectral boundaries of limiter instrument being aimed at the spectral boundaries that HFR produces signal, and the method comprises: frequency boundary HFR being produced signal is added to the border table used when creating frequency band border that limiter instrument uses; And force limiter use the frequency boundary after being added as conservative boundary and correspondingly adjust remaining border.

One embodiment relates to the combination transposition of a sound signal, is included in some the integer transposition rank in a low resolution filter-bank domain, wherein this transposition of time onblock executing operation of subbands signal.

One another embodiment relates to combination transposition, and the transposition rank being wherein greater than 2 are embedded in one 2 rank transposition environment.

One another embodiment relates to combination transposition, and the transposition rank being wherein greater than 3 are embedded in one 3 rank transposition environment, and lower than 4 transposition rank separated perform.

One another embodiment relates to combination transposition, and wherein transposition rank (such as transposition rank are greater than 2) are created by the transposition rank (that is, especially lower-order) previously calculated of copy package containing core encoder bandwidth.Each combination that can expect on available transposition rank and core bandwidth rank is all feasible, instead of restrictive.

The computation complexity that the analysis filterbank decreased number of one embodiment design required for transposition causes reduces.

One embodiment relates to the device producing a bandwidth expansion signal from an input audio signal, this device comprises a patcher, for repairing an input audio signal to obtain one first repair signal and one second repair signal, this second repair signal has different from the first repair signal one and repairs frequency, wherein this first repair signal uses one first patch algorithm to produce, and this second repair signal uses one second patch algorithm to produce; And a combiner, for combining the first repair signal and the second repair signal, to obtain bandwidth expansion signal.

Another embodiment relates to according to aforesaid device, and wherein the first patch algorithm is a harmonic wave patch algorithm, and the second patch algorithm is a non-harmonic patch algorithm.

Another embodiment relates to aforementioned means, and wherein, the first repairing frequency repairs frequency lower than second or vice versa.

Another embodiment relates to aforementioned means, and wherein input signal comprises a repair information; And the repair information that wherein patcher is configured to by extracting from input signal controls, to change the first patch algorithm or the second patch algorithm according to repair information.

Another embodiment relates to aforementioned means, and wherein, patcher is for repairing the block subsequently of audio signal samples, and wherein patcher is configured to the same block the first patch algorithm and the second patch algorithm being applied to audio sample.

Another embodiment relates to aforementioned means, and wherein, patcher comprises a withdrawal device controlled by bandwidth expansion factor, a bank of filters and for the stretcher of bank of filters sub-band signal with random order.

Another embodiment relates to aforementioned means, and stretcher comprises block extraction apparatus, extracts some overlapping blocks for extracting prior value according to one; Phase regulator or window device, for adjusting the subband samples value in each block based on a window function or a phase correction; And overlap/totalizer, the overlap-add process that the overlapping prior value extracting prior value performs window and phase place adjustment block is greater than for using one.

The device that another embodiment relates to for carrying out bandwidth expansion to sound signal comprises: bank of filters, for filtering audio signals to obtain down-sampling sub-band signal; Multiple different sub-band processor, for processing different sub-band signal by different way, this sub-band processor uses different broadening factors to perform different sub-band signal time explanation operations; And combiner, merge for the process sub-band that multiple different sub-band processor is exported to obtain a bandwidth extended audio signal.

Another embodiment relates to a kind of device for carrying out down-sampling to sound signal and comprises a modulator; Use an interpolator of an interpolation factor; One multiple low-pass filter; And one uses one to extract the withdrawal device of the factor, wherein this extraction factor is higher than interpolation factor.

One embodiment relates to a kind of device for carrying out down-sampling to sound signal, comprises: the first bank of filters, and for producing multiple sub-band signal from sound signal, wherein the sampling rate of this sub-band signal is less than the sampling rate of sound signal; At least one synthesis filter banks, after being positioned at analysis filterbank, and for performing sample rate conversion, the number of active lanes that this synthesis filter banks has is different from the number of active lanes of analysis filterbank; Time explanation processor, for the treatment of sample rate switching signal; And combiner, for by time explanation signal and a low band signal or different time broadened signal combination.

Another embodiment is designed for the device by non-integer down-sampling factor down-sampling one sound signal, comprises: a digital filter; One interpolator with an interpolation factor; The one heterogeneous element with odd number and even tap; And one has and is greater than one of interpolation factor and extracts the withdrawal device of the factor, this extraction factor and interpolation factor are selected such that interpolation factor is non-integer with the ratio extracting the factor.

One embodiment relates to a kind of device for the treatment of a sound signal, comprise: core decoder, the one synthesis transform size slight Graph One factor larger than nominal conversion of this core decoder, makes the core decoder be less than corresponding to the nominal sampling rate of nominal conversion size by a sampling rate produce an output signal; And the preprocessor that has one or more bank of filters, one or more time explanation device and a combiner, wherein the filter bank channel number of this one or more bank of filters is few compared to the number determined by nominal conversion size.

Another embodiment relates to a kind of device for the treatment of a low band signal, comprises: one repairs generator, produces multiple repairing for utilizing low band audio signal; Envelope adjuster, for using the envelope for the given scale factor adjustment signal of the adjacent scale factor with scale factor border, wherein this repairing generator is configured for execution and repeatedly repairs, and makes the border between adjacent repairing consistent with the border between scale factor adjacent in frequency marking.

One embodiment relates to a kind of device being used for process one low band audio signal, comprises: repair generator, produce multiple repairing in order to use low band audio signal; And envelope adjustment limiter, for the envelope adjusted value by carrying out limiting a signal in the adjacent limiter frequency band with limiter frequency band border, wherein this repairing generator be configured to perform repeatedly repair, to make the border between adjacent repairing consistent with the border between the adjacent limiter frequency band in a frequency marking.

Process of the present invention is useful for strengthening the audio codec depending on bandwidth extension schemes, and especially, if be high-importance at the perceptual quality of given next the best of bit rate, and process electric power is a restricted resource simultaneously.

The most outstanding application is audio decoder, to be usually embodied on hand-held device and thus with a battery-powered operations.

Coding audio signal of the present invention can be stored on a digital storage media, or can be transmitted on a transmission medium (such as the Internet) of such as a wireless medium or a wire transmission medium.

According to specifically implementing demand, embodiments of the present invention can to implement in hardware or software.Enforcement can utilize a digital storage media to perform, for example, one floppy disk, a DVD, a CD, ROM, a PROM, EPROM, an EEPROM or flash memory, it stores electronically readable control signal, it cooperates with a programmable computer system (maybe can cooperate), to make it possible to perform each method.

Comprise the data carrier that has electronically readable control signal according to certain embodiments of the present invention, this control signal can cooperate with a programmable computer system, makes it possible to perform one of all methods described herein.

Usually, embodiments of the present invention can be implemented to the computer program that has program code, and this program code is used for the one performed when computer program runs on a computer in all methods.Program code such as can be stored in a machine-readable carrier.

Other embodiment comprise for perform method described herein, the computer program be stored in a machine-readable carrier.

In other words, therefore an embodiment of the inventive method is a computer program with program code, and when computer program runs on a computer, this program code one of to be used for performing in all modes as herein described.

Therefore the another embodiment of the inventive method is a data carrier (or a digital storage media, or a computer-readable media), comprises record thereon for performing the computer program of one of all methods described herein.

Therefore another embodiment of the inventive method is that an expression is for performing data stream or a burst of the computer program of one of method described herein.Data stream or burst such as can be configured to connect (such as via the Internet) via a data communication and be transmitted.

Another embodiment comprises a treatment facility, such as a computing machine or a programmable logical device, and this logical device is configured or for one of performing in all methods described herein.

Another embodiment comprises the computing machine it being provided with the computer program for performing one of all methods described herein.

In some embodiments, a programmable logic device (PLD) (such as field programmable gate array) can be used to some or all functions of performing in method described herein.In some embodiments, a field programmable gate array can cooperate with a microprocessor to perform one of all methods described herein.Usually, method is preferably performed by arbitrary hardware unit.

Above-mentioned embodiment, only for illustration of principle of the present invention, should be appreciated that the amendment of configuration described herein and details and modification are apparent for those skilled in the art.Therefore, mean only to be limited by Patent right requirement subsequently, instead of limited by the detail provided in the mode of the description of embodiment herein and explanation.

Document:

[1]M.Dietz,L.Liljeryd,K. and O.Kunz,“Spectral Band Replication,anovel approach in audio coding,”in 112th AES Convention,Munich,May 2002.

[2]S.Meltzer,R. and F.Henn,“SBR enhanced audio codecs for digitalbroadcasting such as“Digital Radio Mondiale”(DRM),”in 112th AESConvention,Munich,May 2002.

[3]T.Ziegler,A.Ehret,P.Ekstrand and M.Lutzky,“Enhancing mp3with SBR:Features and Capabilities of the new mp3PRO Algorithm,”in 112th AESConvention,Munich,May 2002.

[4]International Standard ISO/IEC 14496‐3:2001/FPDAM 1,“BandwidthExtension,”ISO/IEC,2002.Speech bandwidth extension method andapparatus Vasu Iyengar et al

[5]E.Larsen,R.M.Aarts,and M.Danessis.Efficient high‐frequencybandwidth extension of music and speech.In AES 112th Convention,Munich,Germany,May 2002.

[6]R.M.Aarts,E.Larsen,and O.Ouweltjes.A unified approach to low‐andhigh frequency bandwidth extension.In AES 115th Convention,New York,USA,October 2003.

[7]K. A Robust Wideband Enhancement for Narrowband SpeechSignal.Research Report,Helsinki University of Technology,Laboratory ofAcoustics and Audio Signal Processing,2001.

[8]E.Larsen and R.M.Aarts.Audio Bandwidth Extension‐Application topsychoacoustics,Signal Processing and Loudspeaker Design.John Wiley &Sons,Ltd,2004.

[9]E.Larsen,R.M.Aarts,and M.Danessis.Efficient high‐frequencybandwidth extension of music and speech.In AES 112th Convention,Munich,Germany,May 2002.

[10]J.Makhoul.Spectral Analysis of Speech by Linear Prediction.IEEETransactions on Audio and Electroacoustics,AU‐21(3),June 1973.

[11]United States Patent Application 08/951,029,Ohmori,et al.Audioband width extending system and method

[12]United States Patent 6895375,Malah,D&Cox,R.V.:System forbandwidth extension of Narrow‐band speech

[13]Frederik Nagel,Sascha Disch,“A harmonic bandwidth extensionmethod for audio codecs,”ICASSP International Conference on Acoustics,Speech and Signal Processing,IEEE CNF,Taipei,Taiwan,April 2009

[14]Frederik Nagel,Sascha Disch,Nikolaus Rettelbach,“A phase vocoderdriven bandwidth extension method with novel transient handling for audiocodecs,”126th AES Convention,Munich,Germany,May 2009

[15]M.Puckette.Phase‐locked Vocoder.IEEE ASSP Conference onApplications of Signal Processing to Audio and Acoustics,Mohonk 1995.", A.:Transient detection and preservation in the phase vocoder；citeseer.ist.psu.edu/679246.html

[16]Laroche L.,Dolson M.:“Improved phase vocoder timescalemodification of audio",IEEE Trans.Speech and Audio Processing,vol.7,no.3,pp.323‐‐332,

[17]United States Patent 6549884Laroche,J.&Dolson,M.:Phase‐vocoder pitch‐shifting

[18]Herre,J.；Faller,C.；Ertel,C.；Hilpert,J.； A.；Spenger,C,“MP3Surround:Efficient and Compatible Coding of Multi‐Channel Audio,”116thConv.Aud.Eng.Soc.,May 2004

[19]Neuendorf,Max；Gournay,Philippe；Multrus,Markus；Lecomte,Jérémie；Bessette,Bruno；Geiger,Ralf；Bayer,Stefan；Fuchs,Guillaume；Hilpert,Johannes；Rettelbach,Nikolaus；Salami,Redwan；Schuller,Gerald；Lefebvre,Roch；Grill,Bernhard:Unified Speech and Audio Coding Scheme for HighQuality at Lowbitrates,ICASSP 2009,April 19‐24,2009,Taipei,Taiwan

[20]Bayer,Stefan；Bessette,Bruno；Fuchs,Guillaume；Geiger,Ralf；Gournay,Philippe；Grill,Bernhard；Hilpert,Johannes；Lecomte,Jérémie；Lefebvre,Roch；Multrus,Markus；Nagel,Frederik；Neuendorf,Max；Rettelbach,Nikolaus；Robilliard,Julien；Salami,Redwan；Schuller,Gerald:A Novel Scheme for LowBitrate Unified Speech and Audio Coding,126th AES Convention,May 7,2009,

Accompanying drawing explanation

Now, describe the present invention with reference to the accompanying drawings in the mode of illustrated examples, illustrated examples does not limit, in accompanying drawing scope and spirit of the present invention:

Fig. 1 shows and use one of 2,3 and 4 rank transposition based on the operation of the transposition device of block in a HFR enhanced decoder framework;

Fig. 2 shows the operation of the nonlinearities electric band spread unit in Fig. 1;

Fig. 3 shows the efficient enforcement of the transposition device based on block of Fig. 1, wherein uses multi tate time domain resampler and implements resampler before HFR analysis filterbank and bandpass filter based on the bandpass filter of QMF;

Fig. 4 shows the example of a structure block of the multi tate time domain resampler for efficiently implementing Fig. 3;

Fig. 5 shows for the effect of 2 rank transposition by a signal example of the different masses process in Fig. 4;

Fig. 6 shows the efficient enforcement of the transposition device based on block of Fig. 1, and the resampler wherein before HFR analysis filterbank and bandpass filter are replaced with the little sub sampling synthesis filter banks to the sub-band operation selected from one 32 ?frequency range analysis bank of filters;

Fig. 7 shows the effect of the example signal for the sub sampling synthesis filter banks process in 2 rank transposition Fig. 6;

Fig. 8 shows the enforcement block of the efficient multi tate time domain down-sampler of Graph One factor 2;

Fig. 9 shows the enforcement block of the efficient multi tate time domain down-sampler of Graph One factor 3/2;

The spectral boundaries that Figure 10 shows HFR transposition device signal in a HFR enhancement mode scrambler adjusts aiming at of the border of frequency band with envelope;

Figure 11 shows the situation occurring artifact due to out-of-alignment HFR transposition device signal spectrum border;

Figure 12 shows the situation avoiding the artifact of Figure 11 due to the aligning spectral boundaries of HFR transposition device signal;

Figure 13 shows the adjustment of the spectral boundaries in limiter instrument to the spectral boundaries of HFR transposition device signal;

Figure 14 shows the principle of the harmonic wave transposition based on sub-band block;

Figure 15 shows and uses several rank of transposition to apply the illustrative case of the transposition based on sub-band block in a HFR enhancement mode audio coder;

Figure 16 shows the prior art illustrative case of the operation of the transposition based on multistage sub-band block of the analysis filterbank be separated for each transposition rank application one;

Figure 17 shows the single 64 frequency band QMF analysis filterbank of application one and carries out one based on an invention illustrative case of the efficient operation of the transposition of multistage sub-band block;

Figure 18 shows another example for the formation of the process of sub-band signal wisdom;

Figure 19 shows single-sideband modulation (SSB) and repairs;

Figure 20 shows a harmonic wave bandwidth expansion (HBE) and repairs;

Figure 21 shows a mixing and repairs, and first to repair be to be produced by frequency expansion and second to repair be produced by the SSB of low frequency part copy;

Figure 22 shows and utilizes a HBE to repair to produce the one second selectable mixing repairing of repairing to a SSB copy function;

Figure 23 shows the preferred cascade structure analyzed with synthesis filter banks;

Figure 24 a shows one of the little synthesis filter banks of Figure 23 and preferably implements;

Figure 24 b shows one of another analysis filterbank of Figure 23 and preferably implements;

Figure 25 a show ISO/IEC 14496 ?certain of 3:2005 (E) analyze the overview with synthesis filter banks, and the particularly enforcement of an analysis filterbank that uses of an analysis filterbank that can be Figure 23, and the enforcement of a synthesis filter banks that the last synthesis filter banks that can be Figure 23 uses;

Figure 25 b shows the enforcement of the process flow diagram of the analysis filterbank as Figure 25 a;

Figure 25 c shows one of the synthesis filter banks of Figure 25 a and preferably implements;

Figure 26 shows the overview of framework under a bandwidth expansion process background; And

Figure 27 shows preferably implementing of being exported by another analysis filterbank process sub-band signal of Figure 23.

Claims (21)

1., for the treatment of a device for an input audio signal (2300), comprise:

One synthesis filter banks (2304), in order to synthesize an audio frequency M signal (2306) from described input audio signal (2300), multiple first sub-band signals (2303) that described input audio signal (2300) is produced by an analysis filterbank (2302) represent, the filter bank channel number (M of wherein said synthesis filter banks (2304) _s) be less than the number of active lanes (M) of described analysis filterbank (2302); And

One another analysis filterbank (2307), in order to produce multiple second sub-band signal (2308) from described audio frequency M signal (2306), the number of active lanes (M that wherein said another analysis filterbank (2307) has _a) different from the number of active lanes of described synthesis filter banks (2304), make the sampling rate of the sub-band signal in described multiple second sub-band signal (2308) different from the sampling rate of one first sub-band signal in described multiple first sub-band signal (2303).

2. device according to claim 1, wherein, described synthesis filter banks (2304) is a real-valued bank of filters.

3. device according to claim 1, wherein, the number of the first sub-band signal in described multiple first sub-band signal (2302) is more than or equal to 24, and

Wherein, the filter bank channel number of described synthesis filter banks (2304) is less than or equal to 22.

4. device according to claim 1, wherein, described synthesis filter banks (2304) is configured for only process and represents a subgroup (2305) of all first sub-band signals (2303) in described multiple first sub-band signals of full bandwidth input audio signal (2300), and wherein, described synthesis filter banks (2304) is configured for and produces described audio frequency M signal (2306) as the band segments being modulated to base band in described full bandwidth input audio signal (2300).

5. device according to claim 1, comprises further:

Described analysis filterbank (2302) for receive described input audio signal (2300) a time-domain representation and for analyzing described time-domain representation to obtain described multiple first sub-band signal (2303), wherein, one subgroup (2305) of described multiple first sub-band signal (2303) is input in described synthesis filter banks (2304), and wherein, the residue sub-band signal in described multiple first sub-band signal is not input in described synthesis filter banks (2304).

6. device according to claim 1, wherein, described analysis filterbank (2302) is a complex value bank of filters, wherein, described synthesis filter banks (2304) comprises one for calculating the real-valued counter of real-valued sub-band signal from described first sub-band signal, wherein, the described real-valued sub-band signal calculated by described real-valued counter is processed by described synthesis filter banks (2304) to obtain described audio frequency M signal (2306) further.

7. device according to claim 1, wherein, described another analysis filterbank (2307) is a complex value bank of filters and is configured to produce described multiple second sub-band signal (2308) as multiple sub-band signal.

8. device according to claim 1, wherein, described synthesis filter banks (2304), another analysis filterbank described (2307) or described analysis filterbank (2302) are configured to the sub-sampled version using same bank of filters window.

9. device according to claim 1, comprises further:

One sub-band signal processor (2309), for the treatment of described multiple second sub-band (2308); And

One another synthesis filter banks (2311), for carrying out filtering to multiple processed sub-band, wherein, described another synthesis filter banks (2311), described synthesis filter banks (2304), described analysis filterbank (2302) or another analysis filterbank described (2307) are configured to the sub-sampled version using same bank of filters window, or wherein said another synthesis filter banks (2311) is configured to application one and synthesizes window, and wherein said another analysis filterbank (2307), described synthesis filter banks (2304) or described analysis filterbank (2302) are configured to the sub-sampled version applying the described synthesis window that described another synthesis filter banks (2311) uses.

10. device according to claim 1, comprises further and performs Nonlinear Processing operation with the sub-band processor (2309) obtaining multiple processed sub-band for every sub-band;

One high-frequency reconstruction processor (1030), in order to adjust an input signal according to the parameter (1040) of transmission; And

One another synthesis filter banks (2311,1050), in order to combine described input audio signal (2300) and described multiple processed sub-band signal;

Wherein, described high-frequency reconstruction processor (1030) is configured to process another synthesis filter banks (1050 described, 2311) a output, or multiple processed sub-band described in the pre-treatment of another synthesis filter banks (1030) described in being input at described multiple processed sub-band.

11. devices according to claim 1, wherein, described another analysis filterbank (2307) or described synthesis filter banks (2304) have a prototype window function counter, described prototype window function counter is by utilizing the information of the number of active lanes about described another analysis filterbank (2307) or described composite wave device group (2304), use the memory window function with the bank of filters of a different size to carry out sub sampling or interpolation, calculate a prototype window function.

12. devices according to claim 1, wherein, the input of described synthesis filter banks (2304) the lowest filter group passage and a most high filter group passage that are configured near described synthesis filter banks (2304) is set as zero.

13. devices according to claim 1, are configured to the harmonic wave transposition of execution one based on block, and wherein said synthesis filter banks (2304) is a sub sampling bank of filters.

14. devices according to claim 1, comprise the sub-band processor (2309) for the treatment of described multiple second sub-band (2308) further,

Wherein, described sub-band processor (2309,1020a, 1020b) comprises a withdrawal device controlled by bandwidth expansion factor with random order; And for the stretcher of a sub-band signal, wherein said stretcher comprises a block extraction apparatus, extract multiple overlapping block in order to extract prior value according to one; One phase regulator (1806,124a, 124b, 124c) or window device (1802,126a, 126b, 126c), in order to adjust the sub-band sampling in each block based on a window function or a phase correction; And an overlapping totalizer (1804,130), in order to the overlap-add process using an overlapping prior value being greater than described extraction prior value to perform window and phase place adjustment block.

15. devices according to claim 1, comprise a sub-band processor (2309) further, and wherein said sub-band processor (2309,1020a, 1020b) comprises:

For multiple different disposal branch roads (110a, 110b, 110c) of the different transposition factor, for obtaining a transposition signal, wherein, each process branch road is configured to the block extracting (120a, 120b, 120c) subband samples;

One totalizer (128), for by described transposition signal plus to obtain transposition block; And

One overlapping totalizer (130), in order to use ratio at described multiple different disposal branch road (110a, 110b, (120a is extracted 110c), 120b, the block prior value that 120c) the block prior value that uses of block is large, the transposition block of overlap-add Time Continuous.

16. devices according to claim 1, comprise further:

Described analysis filterbank (2302), wherein said synthesis filter banks (2304) and another analysis filterbank described (2307) are configured to the conversion of execution one sample rate,

One time explanation processor (100a, 100b, 100c), in order to process the signal of sample rate conversion; And

One combiner (2311,605), the treated sub-band signal produced by described time explanation processor in order to combination, to obtain a treated time-domain signal.

17. devices according to claim 1, wherein, the number of active lanes of described another analysis filterbank (2307) is greater than the number of active lanes of described synthesis filter banks (2304).

The device of 18. 1 kinds of process one input audio signal (2300), comprises:

One analysis filterbank (2302), there is multiple (M) analysis filterbank passage, wherein, described analysis filterbank (2302) is configured to carry out filtering to described input audio signal (2300), to obtain multiple first sub-band signal (2303); And

One synthesis filter banks (2304), an audio frequency M signal (2306) is synthesized for using the group of the first sub-band signal (2303) (2305), wherein, described group (2305) comprise the sub-band signal that a number is less than the filter bank channel number of described analysis filterbank (2302), and wherein said middle sound signal (2306) is that the sub sampling of a portions of bandwidth of described input audio signal (2300) represents.

19. devices according to claim 18, wherein, described analysis filterbank (2302) is the multiple QMF bank of filters of threshold sampling, and

Wherein, described synthesis filter banks (2304) is the real-valued QMF bank of filters of a threshold sampling.

The method of 20. 1 kinds of process one input audio signal (2300), comprises:

The synthesis filter banks (2304) being used for synthesizing from described input audio signal (2300) an audio frequency M signal (2306) is used to carry out synthetic filtering to the described input audio signal (2300) that multiple first sub-band signals (2303) produced by an analysis filterbank (2302) represent, wherein, the number of active lanes (M of the bank of filters of described synthesis filter banks (2304) _s) be less than the number of active lanes (M) of described analysis filterbank (2302); And

Another analysis filterbank (2307) being used for producing from described audio frequency M signal (2306) multiple second sub-band signal (2308) is used to carry out analysis filtered, the number of active lanes (M that wherein said another analysis filterbank (2307) has _a) be different from the number of active lanes of described synthesis filter banks (2304), make the sampling rate of a sub-band signal of described multiple second sub-band signal (2308) be different from the sampling rate of one first sub-band signal of described multiple first sub-band signal (2303).

21. 1 kinds, for the treatment of the method for an input audio signal (2300), comprise:

Use an analysis filterbank (2302) with multiple (M) analysis filterbank passage to carry out analysis filtered, wherein said analysis filterbank (2302) is configured to input audio signal described in filtering (2300) to obtain multiple first sub-band signal (2303); And

The synthesis filter banks (2304) that use is used for utilizing the group of the first sub-band signal (2303) (2305) to synthesize an audio frequency M signal (2306) carries out synthetic filtering, wherein, the number of the described group of sub-band signal comprised is less than the filter bank channel number of described analysis filterbank (2302), wherein, described middle sound signal (2306) is that the sub sampling of a portions of bandwidth of described input audio signal (2300) represents.